mlsanityData to 14 Oct 2026

Importance Weighted Expectation-Maximization for Protein Sequence Design

arXiv 2305.00386 ↗First posted 30 Apr 2023q-bio.BM cs.LGcited by 12

Not on the map. Why?

Abstract (arXiv)

Designing protein sequences with desired biological function is crucial in biology and chemistry. Recent machine learning methods use a surrogate sequence-function model to replace the expensive wet-lab validation. How can we efficiently generate diverse and novel protein sequences with high fitness? In this paper, we propose IsEM-Pro, an approach to generate protein sequences towards a given fitness criterion. At its core, IsEM-Pro is a latent generative model, augmented by combinatorial structure features from a separately learned Markov random fields (MRFs). We develop an Monte Carlo Expectation-Maximization method (MCEM) to learn the model. During inference, sampling from its latent space enhances diversity while its MRFs features guide the exploration in high fitness regions. Experiments on eight protein sequence design tasks show that our IsEM-Pro outperforms the previous best methods by at least 55% on average fitness score and generates more diverse and novel protein sequences.

Key facts. Cited by 12 papers in the map (data through 14 Oct 2026). Builds on 17 of its 77 references that are in the map. First posted on arXiv 30 Apr 2023.

Builds on

17 of 77 references.

References · 17

  1. 2017Attention Is All You Needcited by 60,361
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2019Evaluating Protein Transfer Learning with TAPEcited by 231
  4. 2022Biological Sequence Design with GFlowNetscited by 174
  5. 2020ProGen: Language Modeling for Protein Generationcited by 128
  6. 2020Optimus: Organizing Sentences via Pre-trained Modeling of a Latent Spacecited by 114
  7. 2019Conditioning by adaptive sampling for robust designcited by 111
  8. 2020Glancing Transformer for Non-Autoregressive Neural Machine Translationcited by 103
  9. 2022Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimizationcited by 97
  10. 2018Iterative Amortized Inferencecited by 94
  11. 2019Model Inversion Networks for Model-Based Optimizationcited by 83
  12. 2020AdaLead: A simple and robust adaptive greedy search algorithm for sequence designcited by 65
5 more
  1. 2020Optimizing Molecules using Efficient Queries from Property Evaluationscited by 50
  2. 2018Design by adaptive samplingcited by 32
  3. 2021Unifying Likelihood-free Inference with Black-box Optimization and Beyondcited by 11
  4. 2019Reweighted Expectation Maximizationcited by 9
  5. 2021Benchmarking deep generative models for diverse antibody sequence designcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 3; 2024: 5; 2025: 3; 2026: 1.

12 papers.

  1. 2023Improving Protein Optimization with Smoothed Fitness Landscapescited by 21
  2. 2024SurfPro: Functional Protein Design Based on Continuous Surfacecited by 14
  3. 2024Improved Off-policy Reinforcement Learning in Biological Sequence Designcited by 13
  4. 2025Steering Generative Models with Experimental Data for Protein Fitness Optimizationcited by 7
  5. 2024Generative Enzyme Design Guided by Functionally Important Sites and Small-Molecule Substratescited by 5
  6. 2024Cell Morphology-Guided Small Molecule Generation with GFlowNetscited by 4
  7. 2023Joint Design of Protein Sequence and Structure based on Motifscited by 3
  8. 2025Overconfident Oracles: Limitations of In Silico Sequence Design Benchmarkingcited by 3
  9. 2025ProSpero: Active Learning for Robust Protein Design Beyond Wild-Type Neighborhoodscited by 3
  10. 2023Functional Geometry Guided Protein Sequence and Backbone Structure Co-Designcited by 2
  11. 2024TaxDiff: Taxonomic-Guided Diffusion Model for Protein Sequence Generationcited by 0
  12. 2026Efficient Protein Optimization via Structure-aware Hamiltonian Dynamicscited by 0